Server Hardware Explained (Part 6)
This article continues the series on server hardware by discussing solid state hard drives.
In my previous article I began talking aboutÂ server storage and I spent quite a bit of time talking about the differences between consumer grade SATA drives and the SATA drives that are typically sold for use with servers. Of course when it comes to server storage, SATA is not your only option. SAS drives and solid state drives are also popular choices. In this article, I want to turn my attention to solid state drives. As I do, I will discuss the advantages and disadvantages of using solid state drives as well as a way to get the best of both worlds.
Solid State Drives
Solid State Drives (SSDs) have been around for several years, but have only recently begun to gain mainstream acceptance. SSDs use semiconductor memory such as NAND flash memory instead of the magnetic platters found in traditional hard drives. The memory used in SSDs is very similar to the type of memory found in USB flash drives and in SD cards.
SSD drives offer a number of advantages over traditional hard drives (HDDs). Most of these advantages are a direct result of the fact that unlike a HDD, an SSD does not contain any moving parts.
Because there are no moving parts in SSDs, they consume less power than traditional HDDs. According to Intel, SSDs use 20% less power than HDDs. Furthermore, because less power is being used (and because there are no moving parts generating friction), SSDs give off far less heat than HDDs.
Another nice benefit to having a drive with no moving parts is that SSDs operate silently. Granted, most HDDs arenât really all that loud but if you have a server with dozens of HDDs the combined noise can be considerable.
While these benefits are nice, the main advantages to using SSDs are that they offer far better performance and reliability than HDDs. The reason why SSDs offer such high performance is because the driveâs speed is not limited by mechanical components. The operating system will never have to wait for a SSD drive to spin up and there is no waiting for a mechanical drive head to move across the platter to the required location.
This goes a long way toward improving the speed of random access. In fact, Intel did some benchmark tests in which they compared the speed of an Intel X25-M SSD against a standard SATA drive. In one test they measured how long it took Windows to boot off of each drive. The boot process took 30 seconds on the HDD and was reduced to 19 seconds on the SSD. Additional benchmark testing concluded that the maximum transfer rate for an SSD was more than double that of an HDD for both read and write operations.
SSDs also tend to be more reliable than HDDs. Because there are no moving parts there arenât any mechanical mechanisms that can wear out. Likewise, SSDs are less likely than HDDs to break if they are dropped. According to Intel, SSDs have roughly double the Mean Time Between Failure (MTBF) of HDDs. In other words, a SSD should theoretically last twice as long as a HDD.
In spite of the many advantages to using solid state drives, there are at least two major disadvantages that must be taken into account â price and capacity. At the time that this article was written (November of 2011) SSDs offered far less capacity and were far more expensive than their HDD counterparts.
By way of comparison, the largest SATA III compatible SSD that I could find had a mere 512 GB capacity and sold for $849 (US Dollars). By way of comparison the same Web site offered a standard 2.5 inch SATA drive with a comparable capacity for $109. Furthermore, while the maximum size of a 2.5 inch solid state drive was half a gigabyte, the Web site offered 2.5 inch HDDs with a capacity of 1.5 TB. Of course some servers use 3.5 inch drives, which greatly increases the potential capacity since some manufacturers currently offer drives that hold up to 3 TB. The point is that if you want to use solid state drives you will have to sacrifice capacity and pay a premium price (at least for now).
Most of the SSDs on the market use a SATA interface.Â This allows SSDs to be used in any computer that has a compatible SATA controller. However, the SATA interface was originally designed for use with HDDs and it is the SATA interface that is usually the performance bottleneck when SSDs are used. Please donât misunderstand me. All of the benefits that were previously discussed can be achieved even if the SSD has a SATA interface. However, the drives themselves are actually capable of greater performance than what SATA can deliver. This has led to the creation of hybrid SSDs.
Hybrid SSDs arenât widely used on servers, but they might be in the not too distant future, so I wanted to briefly mention them. A hybrid SSD does not use a SATA interface. Instead, the drive is mounted directly on a PCIe card. This allows for much greater performance since the PCIe bus is faster than the SATA bus.
The reason why hybrid SSDs are called hybrid drives is because the PCI card makes use of both SSD and HDD storage. The basic idea is that data which is frequently accessed is stored on the SSD portion of the drive so that it can be retrieved quickly. Data that is not used as often is written to a standard HDD that is mounted directly onto the PCIe card. This design not only offers high performance, but it also allows for greater storage capacity than what is currently possible with a SSD drive. While researching this article, I found a 1 TB hybrid drive for $499. In case you are wondering, this particular model included 100 GB of SSD storage and just under 1 TB of HDD storage.
Although SSDs come at a price per gigabyte that far exceeds that of more traditional HDD storage, using solid state storage is not always cost prohibitive. Some servers simply do not require high capacity storage. For instance Web servers, domain controllers, DNS Servers, DHCP Servers and other types of infrastructure servers generally require relatively little disk space.Â Such servers almost always use less than 100 GB of storage space, and 100 GB SSDs are available for under $200.
In case you are wondering, SSDs can be configured in an array just as HDDs can. It is therefore possible to perform striping, mirroring, and to build other storage architectures using SSDs.
Now that I have discussed solid state storage, I want to turn my attention to SAS drives, which are less expensive and more common than solid state drives. I will discuss SAS storage options in Part 7.
We have 31 guests and no members online
Articles Most Read